Starting in the late 1990s, a potent political divide emerged in the United States. White rural Americans, who previously had voted similarly to white urban dwellers in presidential elections, pulled away from them and became increasingly strong and steadfast supporters of Republican Party candidates.Footnote 1 Many other nations have long featured rural-urban political cleavages, but the two American political parties historically each gained support from both more and less populated places. Until roughly a quarter-century ago, many rural areas in the United States were electorally competitive and where one party dominated, it was at least as likely to be the Democratic Party as the Republican. Since then, the rural-urban political divide has manifested itself in all regions of the nation, exacerbating social and political polarization (Brown, Mettler, and Puzzi Reference Brown, Mettler and Puzzi2021). Because of the nation’s relatively unique political institutions, moreover, this development has also given the Republican Party more influence than its electoral support would otherwise allow (Rodden Reference Rodden2019).
Recently, political scientists studying the United States have begun to expose this place-based cleavage and to describe it in rich detail (e.g., McKee Reference McKee2008). Most research to date has focused on how individual rural people see or think about their circumstances and politics, stressing their distinct rural consciousness, identity, or values (e.g., Cramer Reference Cramer2016; Gimpel et al. Reference Gimpel, Lovin, Moy and Reeves2020; Lunz Trujillo Reference Lunz Trujillo2022; Nelsen and Petsko Reference Nelsen and Petsko2021; Diamond Reference Diamond2023). This literature has successfully placed the rural-urban divide on political scientists’ research agenda, revealed it to be a key dimension of contemporary political polarization, and generated several important observations that we use to help generate hypotheses. However, scholars have yet to convincingly address the questions of why the rural-urban political divide emerged nationwide when it did and why it has continued to grow (for an exception, see Rodden Reference Rodden2019; on the South, see Hood and McKee Reference Hood and McKee2022).
In drawing attention to these empirical gaps, we are effectively contributing to literature that examines why a political cleavage that did not previously exist may emerge at a particular time and then become deeply entrenched. Society always contains differences between groups, but only some may have political significance, and which ones do may vary. In some instances, for example, religion or ethnicity foster political polarization, whereas in others, they do not, and class distinctions may instead sort citizens between the parties. Why then, over the course of just a few decades, did the United States transition from lacking a partisan cleavage between rural and urban areas to gaining one, and why did it become so entrenched?
To answer these questions, we build on theories developed by scholars of comparative politics to explain the relationship between social cleavages and political parties. In particular, we draw on Seymour Martin Lipset and Stein Rokkan’s (Reference Lipset and Rokkan1967) framework that stresses both political-economic transformation as well as the nationalization of state development. We bring a developmental analytical lens to bear, taking a longer view than existing studies by exploring patterns in rural and urban politics since the 1970s. This permits us to reveal the timing and sequencing of specific processes that contributed to the emergence of the rural-urban divide. We show that over time, multiple forces each played a role in facilitating a deep cleavage.
Descriptively we show that while rural and urban counties resembled each other in several social, economic, and demographic respects in the 1970s, since that time they have moved apart dramatically. Then we examine how such trends relate to political change in presidential voting. We find that through the 1990s and early 2000s, rural dwellers in places experiencing population loss and economic stagnation began to support Republican candidates. Then in the late 2000s, those in areas with higher percentages of less-educated residents or a high presence of evangelical congregations per capita, both of which are more prevalent in rural areas than urban areas, shifted their support to Republicans. Similarly, higher levels of anti-Black racism among rural whites, particularly in the 2010s, helped consolidate Republican dominance. Through sequential processes of polarization, with political-economic forces leading the way and spurring rural resistance to nationalized policy goals and state development, the rural-urban political divide emerged as a major fault line in the nation’s politics.
Growing Political Divergence across Place and Time
Figure 1 displays, in presidential elections, the share of all votes going to Republican candidates, from all rural and urban counties nationwide, using the Office of Management and Budget’s (OMB) definition of rural and urban, which is widely used by the U.S. Department of Agriculture (USDA). As described in more detail later (see Methods and Data), we use this measure because it captures not only population density but also economic and social integration among counties.Footnote 2 From the 1970s into the 1990s, rural and urban dwellers tended to shift in tandem, supporting particular presidential candidates at very similar rates, with rural people just barely more likely than their urban counterparts to support Republicans. Both regions offered strong support for Republican candidates Nixon and Reagan, the latter particularly in his re-election campaign in 1984, and both areas supported Democrat Bill Clinton at high rates. But from 2000 onward, a stark divide emerged, as rural people increasingly supported the Republican candidate in each election, and urban people, the Democratic candidate. The gulf between them has grown from just two percentage points as recently as 1992 to 21 by 2020.Footnote 3
The rural-urban divide in presidential voting
The rural-urban political divide has widened, furthermore, in all regions of the nation, as shown in figure 2. Certainly it is well known that dramatic political change has occurred in the South, where counties transitioned from overwhelming support for Democrats earlier in the twentieth century to growing support for Republicans. Less well known is that as the South changed, particularly just since 1996, it acquired a striking divide between rural and urban counties, a gap that reached 20 percentage points in the 2020 presidential election (Hood and McKee Reference Hood and McKee2022). Each region of the country, furthermore, experienced this divergence over the same period of time. The Midwest now features an even larger gap than the South, 21 percentage points as of 2020. The West is not far behind, with a gap of 19 percentage points. The rural-urban divide in the Northeast was already a bit greater than in these other regions in 1970, and it has increased less than the others, though the gap is still substantial, at 14 points.
The rural-urban divide in presidential voting, by region
These regional shifts in presidential voting are only one manifestation of this new geographic cleavage. The rural-urban divide in presidential voting has also widened in nearly all states over this same time period. Several predominantly rural states have shifted their support to the Republican Party at multiple levels of governance, furthermore, increasingly electing GOP officials to state houses, governorships, and congressional delegations. We will focus here, though, on vote choice in presidential elections, primarily using counties as our level of analysis. Although our empirical approach highlights developments among both rural and urban Americans, we center our analytical attention on rural areas and dwellers in particular because it is among them that the most sweeping political change has occurred over recent decades. And even though rural people make up a shrinking share of the electorate (see figure 4), the United States’ unique political institutions grant them disproportionate power (Rodden Reference Rodden2019), meaning their shift to more consolidated support for the Republican Party is particularly consequential.
Toward a Developmental Framework
This over-time development of place-based trends in American politics remains largely unexplained in the growing political science literature on the rural-urban divide. Drawing on both ethnographic research and large-N survey analyses, scholars have offered several important findings, including that rural people feel resentful towards urbanites (Jacobs and Munis Reference Jacobs and Munis2022; Cramer Reference Cramer2016); anti-Black racism plays a role in shaping white rural consciousness (Nelsen and Petsko Reference Nelsen and Petsko2021); and rural dwellers tend to support the Republican Party, even after controlling for various otherwise important individual-level covariates (Scala and Johnson Reference Scala and Johnson2017). Others have pointed to politically-based sorting as one potential mechanism (e.g., Bishop Reference Bishop2008; Gimpel and Hui Reference Gimpel and Hui2015), yet recent analysis that draws on fine-grained publicly available voter registration data suggests that this can only explain a fraction of the divergence we outline earlier (Martin and Webster Reference Martin and Webster2020).Footnote 4 Overall, by using primarily data only from recent years and focusing on current attitudes and behavior, this literature has three major limitations that prevent us from understanding when and why any of these factors may have become politically consequential.
First, the focus of this research implies, if inadvertently, that the rural-urban partisan divide in presidential voting is a permanent or long-standing fixture of American politics, when in fact it actually evolved only in the past quarter-century. Certainly over the course of history, sociocultural and political differences often separated rural and urban areas. The key point, however, is that when political distinctions occurred, they varied by region, rather than being uniform nationwide. The contemporary rural-urban divide in presidential voting is distinct in its widespread, national characteristics—it features the same pattern in all regions of the nation—and its enduring qualities. This sweeping transition begs the question of why it developed and when it did. Second, and relatedly, existing research seems to suggest the divide emanates from innate characteristics of rural people, when in fact in voting in presidential elections they acted more similarly to urban dwellers just a few decades ago. Explanations that focus on contemporary attitudes, for example, need to account for why any such factors did or did not have a distinct bearing on rural politics in the not-too-distant past, as well as when their impact emerged and why it has varied over time.
Third, reflecting broader shifts in the political science discipline, scholars tend to center their analyses on single causal variables or explanations. Some focus on economic explanations, such as economic despair among the white working class (e.g., Gest Reference Gest2016). Others show the impact of status threat, identity, racism, or nativism (e.g. Mutz Reference Mutz2018; Schaffner, McWilliams, and Nteta Reference Schaffner, MacWilliams and Nteta2018; Sides, Tesler, and Vavreck Reference Sides, Tesler and Vavreck2018; Parker and Barreto Reference Parker, Barreto, Lieberman, Mettler and Roberts2021).Footnote 5 While each of these approaches conveys a great deal about proximate causes of developments within a short time frame, this type of analysis can lead to oversimplified “either/or” debates that overlook ways that different forces might be at play with each other, generating political development over a greater span of years. Some factors might have yielded effects at an earlier time, and subsequently facilitated or amplified the political relevance of others, in a sequential pattern. For example, economic changes in a prior period may have stoked resentment, and in turn made other issues—involving the regulation of the environment or social policy—more salient later on.
As such, the developmental framework we advance takes a longer view of politics. Scholars of historical institutionalism have argued that understanding politics may require, as Paul Pierson (Reference Pierson2004) puts it, not a “snapshot” but rather a “moving pictures” approach that is attentive to the historical development of phenomena and their timing and sequencing. In adopting this approach, we aim to probe if and when particular trends and historical processes might have contributed to the rural-urban divide, how they might have affected rural and urban populations at different points in time, and how their development may have spurred subsequent ones (e.g., Orren and Skowronek Reference Orren and Skowronek2004; Thelen and Streeck Reference Thelen and Streeck2005).
Moreover, our developmental framework requires a more comprehensive analysis of politics, considering several dynamics and how they interact rather than focusing only on one. In recent decades, the growing emphasis on causal inference has often meant prizing the isolation of the causal effect of one variable on another. We aim instead to reveal, in Eric Schickler’s words, “the interplay of multiple historical processes over the course of several decades” (Schickler Reference Schickler2016, 17), to consider the impact of various potential determinants of the rural-urban divide, as well as their timing and sequencing relative to each other. We seek to build on the existing literature by examining when and how the multiple factors scholars have highlighted separately may have emerged in shaping the rural-urban political divide.
Theorizing about Place-Based Trends
The United States has been buffeted by vast economic, political, and social transformations over the past several decades, and these have manifested themselves differently across places. Have these changes created a partisan cleavage between rural and urban dwellers, and if so, how? How might their relative timing and interplay with one another have influenced political development?
Our analysis builds on Lipset and Rokkan’s classic insight that two fundamental processes of change may drive partisan divisions, each fostering its own line of cleavage. First, they highlight a political-economic dimension, featuring conflict over resources, products, or benefits in the economy, and which in their analysis—focused on Europe in the late nineteenth and early twentieth century—involved the process of industrialization and the spread of global trade.Footnote 6 We update this to examine the economic transformations that began in the late twentieth century and continue in the present. We draw attention to the rise of post-industrial capitalism, particularly the growing concentration of economic activity in urban areas, in part spurred by ascendance of the “knowledge economy.” We ask when such developments deepened an economic divide along rural and urban lines, and how that cleavage related to political party support.
Second, Lipset and Rokkan point to a nationalizing dimension which spurs resistance by those in remote places to the encroachments of “dominant national elites and their bureaucracies … to the pressures of the centralizing, standardizing, and ‘rationalizing’ machinery of the nation-state” (Lipset and Rokkan Reference Lipset and Rokkan1967, 10).Footnote 7 While they focus on how such conflict emerged following the “widening of the suffrage” (1967, 12), in probing the contemporary United States we consider various manifestations of state power. One involves the gradual incorporation, particularly from the 1960s to the present, of Americans as equal citizens through the extension of national laws and public policies that require states and localities to guarantee rights to all regardless of race, ethnicity, gender, or sexuality. (With respect to the latter, some scholars have suggested that the rising emphasis of so-called “social issues” in national politics might itself explain the rural-urban divide [e.g., Hopkins Reference Hopkins2017], but we consider contestation over them to be only one manifestation of this broader phenomena of nationalization.) Another pertains to the use of state power to address other issues ranging from environmental concerns, such as protecting endangered species or regulating coal or solar production, for example, to gun control or liberalizing immigration. As the Democratic Party came to be perceived as the party promoting national adoption of such policies, particularly by the late 1990s and early 2000s, resistance and alienation ensued among Americans who felt imposed on by outsiders who wielded government power. We will be analyzing whether and when the promotion of nationalization and resistance to it fostered a rural-urban divide. We seek to understand how the rural-urban political divide emerged through what Lipset and Rokkan call “sequential interactions between these two fundamental processes of change” (Lipset and Rokkan Reference Lipset and Rokkan1967, 34).
Although the nationalizing trends gained momentum long before the political-economic transformation, we anticipate that only once political-economic forces created upheaval and decay in rural areas did residents become more amenable to changing which party they favored in national elections. This, in turn, may have subsequently spurred the nationalizing dimension to have become a more effective force in mobilizing them to vote for Republicans. Put differently, once white rural voters felt abandoned by Democrats on political economic issues, they engaged in backlash on other issues as well. They chafed especially at new regulations, which they perceived as being promoted by an emergent national urban elite who failed to understand or respect them. Organizations, furthermore, may have helped cement this cleavage by mobilizing rural voters along the nationalization dimension.Footnote 8 We now examine several developments that likely facilitated these processes. As we do so, we offer a set of expectations for our regression analyses to follow, based on the literature and a rich set of descriptive statistics.
Left Behind
A pervasive stylized fact about the rural-urban divide implies that rural dwellers feel “left behind” by massive economic changes and the harsh consequences they have imposed on rural areas. Katherine Cramer (Reference Cramer2016) and Robert Wuthnow (Reference Wuthnow2018), for example, draw on rich ethnographic research to show this. Cramer (Reference Cramer2016) finds among rural dwellers a distinct consciousness that is underpinned by a sense of distributive injustice and fosters resentment toward those they consider urban elites. But scholars have yet to examine broad national trends to illuminate why such alienation and resentment emerged when they did, and why they might have driven growing allegiance to the Republican Party.
Historically, state development in the United States incorporated rural areas into the nation’s growing economy. In the nineteenth century, for example, the U.S. Department of Agriculture widely distributed seeds to farmers (Carpenter Reference Carpenter2001, 179). The agrarian movement of the late nineteenth century failed to win the election of its favored presidential candidate, William Jennings Bryan, but many of its long-held policy goals came to pass in the early twentieth century (Sanders Reference Sanders1999, 173). The New Deal incorporated rural areas through policies ranging from the Farm Bankruptcy Act to the Rural Electrification Act, and through the creation of the Farm Security Administration and the Tennessee Valley Authority.
Over the past three decades, by contrast, upheavals in the political economy have relegated most rural places to an increasingly marginal status. The United States embraced free trade, first through the enactment of North American Free Trade Act (NAFTA) in 1993, and by the early 2000s, through importing more from China. Industries nationwide, both in cities and many rural areas, especially in the South and Midwest, downsized or shuttered their doors. Yet many urban areas found ways to adapt, embracing the rise of the so-called “knowledge economy,” with its high-paying jobs in the technology sector, as well as the broader proliferation of service-sector jobs. Most rural areas, by contrast, struggled to find a way forward, especially as their population aged (Wuthnow Reference Wuthnow2018) and stagnated in size (discussed later). Their employment losses were compounded by technological changes that decimated many rural jobs in agriculture and extractive industries. In addition, financial deregulation, promoted by some Democrats as well as many Republicans, undercut small local banks that were geographically dispersed throughout the nation and which had long provided credit to local businesses (Barton Reference Barton2022). These trends combined to forge different pathways of economic development—and stagnation—that resulted in vast disparities in job growth and population growth between urban and rural areas (Florida and King Reference Florida and King2019).
Our descriptive analysis reveals the divergence in job growth over time. Figure 3 uses population and employment data from the U.S. Census and Bureau of Economic Analysis merged with the OMB’s definition of rurality to construct a per capita jobs measure. Since the rural population has stagnated (see figure 4), weighting employment by population makes it a relatively conservative measure of economic well-being for non-urban areas. Nevertheless, as is shown, since 2000 rural areas have seen stagnant job growth, while urban areas have continued their economic ascent. Indeed, the per capita job gap between rural and urban areas nationally has more than doubled since 1970. We find that a striking 94% of all the job growth in the nation since 2000 has occurred in urban counties. Over the same time period, a full 47% of rural counties have experienced net job loss, while just 15% of urban ones have. Certainly, employment loss is in part related to demographic changes, such as population stagnation and the aging profile of rural areas, yet all these forces nevertheless signal significant decline in the economic fortunes of rural communities.
Employment per capita, over time
Raw population growth, over time
How might job stagnation or decline in rural areas affect politics? Some scholars argue that deindustrialization has pushed whites to vote, increasingly, for Republican candidates. Although they do not consider rural areas in particular, they show how areas subject to trade vulnerability, whether in the 1990s following the adoption of NAFTA or in the 2000s as trade with China grew, have shifted their political support to the GOP (e.g., Choi et al. Reference Choi, Kuziemko, Washington and Wright2021; Wright Reference Wright2020; Autor et al. Reference Autor, Dorn, Hanson and Majlesi2020). Baccini and Weymouth (Reference Baccini and Weymouth2021) offer evidence that job loss among non-Hispanic whites has activated white identity politics and social status anxieties. While many rural dwellers had long associated with the Democratic Party and seen it as acting in their interests, President Clinton’s shift to supporting free trade and some congressional Democrats’ shift to supporting deregulation may have prompted them to reevaluate their party affiliation (Barton Reference Barton2022). Meanwhile, Democrats have done well in the past two presidential elections in areas in which the bulk of economic activity is located, as demonstrated by descriptive analysis using county-level Gross Domestic Product (GDP) (Muro et al. Reference Muro, Byerly Duke, You and Maxim2020; Muro and Liu Reference Muro and Liu2016).
Rural areas have also encountered population loss or stagnation. Figure 4 graphs the diverging population trends across rural and urban counties, using county-level data from the U.S. Census Bureau with OMB rural-urban identifiers. As is displayed, while urban areas have continued to grow at a fast clip, rural counties overall have stagnated. According to our calculations, the disparity has been greatest during recent decades. From 1990 to 2019, urban counties grew by 36%, compared to rural areas that grew by just 17%. And while a mere 12% of all urban counties lost population over that time period, 41% of all rural ones did.
Population loss or stagnation can also shape political behavior. Some scholars suggest that it may foster grievance politics. Population loss tends to lead to the shuttering of local businesses, a decline in services and gathering places, and in some instances even the closing of schools or hospitals (Shepherd Reference Shepherd2021). Wuthnow describes depopulated places as pervaded by an “unspoken sadness,” and often by “anger and frustration;” a small-town dweller in Kansas explained to him that recently conservative Republicans had become “not only angry but ‘very vocal’ about expressing it” (Wuthnow Reference Wuthnow2018, 48, 54). While few scholars of American politics have analyzed population loss systematically, comparativists have shown that it has subsequently facilitated the rise of populist right-wing parties (e.g., see Dancygier et al. Reference Dancygier, Dehdari, Laitin, Marbach and Vernby2022). As services and economic opportunities have clustered around urban areas, Pauliina Patana demonstrates, for example, that spatial inequality has ensued, making it harder for those who continue to live in economically declining areas to leave them. This generates a feeling of being “stuck” or residentially constrained relative to others, and that is associated with support for populist radical right parties (Patana 2022).
In sum, political-economic developments have deepened a cleavage between the primary and peripheral sectors of the economy that falls along the rural-urban line. We expect that in turn, places that have experienced less job growth or greater population stagnation over time, particularly rural ones, may have shifted toward greater support for Republican candidates. To be clear, it is not that Republicans offered an issue agenda that addressed rural job or population loss more effectively than Democrats did, but rather that once issues such as trade and financial regulation no longer differentiated the parties, rural dwellers became alienated from the Democratic Party and other issues became more salient.
Educational Gap
As circumstances in rural and urban areas have diverged in the ways noted earlier, educational attainment has taken on growing significance. Many young adults leave rural areas to seek educational opportunities, and they typically relocate permanently to urban areas (Wuthnow Reference Wuthnow2018, 56-63). Rural dwellers with less education have become less likely to move to urban areas. While these dynamics may be rooted in political-economic transformation, we treat educational attainment as a distinct phenomenon because evidence suggests that it can independently play a key role in shaping citizens’ views on issues (e.g., see Kitschelt and Rehm Reference Kitschelt and Rehm2019), and, we suggest, their perception of the nationalization of state development.
Using data from the U.S. Census Bureau, figure 5 graphs the geographic divergence in college degree attainment over recent decades. In 1970, rural and urban areas differed little on this score; urban counties surpassed rural ones by only five percentage points on rates of four-year degree attainment among adults 25 and older. Yet in more recent decades, urban areas have “pulled away” from rural areas, and by 2020, 35% of all adults in urban areas had at least four years of college education, while 21% of those in rural areas did.
The rural-urban education divide, over time
The growing educational gap between rural and urban places may affect politics by shaping ideas or values. Iversen and Soskice, for instance, observe that advanced education tends to be associated with the adoption of progressive social values on matters of gender, sexuality, race, and immigration, each of which has been promoted in the contemporary United States primarily by the Democratic Party. This may be compounded by the fact that highly educated people today are more likely to live in diverse and fluid urban areas, working and living alongside others who come from different backgrounds than themselves (Iversen and Soskice Reference Iversen and Soskice2019, 225-226). Meanwhile, the two political parties have diverged on such issues particularly since the 1980s, for example when the Republican Party—which previously had endorsed the Equal Rights Amendment—turned against it and ousted pro-choice officials (Rosenfeld Reference Rosenfeld2018).
Yet education may come into play through other dynamics, as well, particularly if it involves a power struggle between more and less educated groups in society. More so than reflecting large gaps in values, one group might perceive the other as imposing their will on them. Here we are drawing on the expectations of Lipset and Rokkan, discussed earlier. We expect that rural dwellers, who are less likely to have advanced education, may have resented what they viewed as the efforts of highly educated urban Democrats to impose their priorities on the nation, subjecting them to policies in which they felt they had little input.
As such, we expect that urban places with a growing percentage of highly educated residents may be shifting toward support for the Democrats over time, and rural places with fewer such residents, toward the Republicans. In this instance, the transformation of the political economy—as urban places thrived and rural places fell behind—may have prompted a sequential nationalizing cleavage, as those in rural places increasingly understood their political affiliation in distinction to those of urban places, and vice versa.
Organizational Mobilization
Civic and political organizations can also play a crucial role in political developments by highlighting conditions such as those mentioned earlier, creating a narrative that assigns credit or blame for such trends, socializing voters in ways that provide a rationale for their political choices, and mobilizing voters to participate in elections. Scholars have highlighted how several organizations in recent decades have generated support for the Republican Party or its shift to the right (e.g., Skocpol and Williamson Reference Skocpol and Williamson2012; Skocpol and Tervo Reference Skocpol and Tervo2020). These include, for example, religious groups (Margolis Reference Margolis, Lieberman, Mettler and Roberts2021), the Koch-affiliated organizations such as Americans for Prosperity (Hertel-Fernandez Reference Hertel-Fernandez2019), gun groups loosely connected to the National Rifle Association (Lacombe Reference Lacombe, Lieberman, Mettler and Roberts2021), and chapters of the Fraternal Order of Police (Zoorab and Skocpol Reference Zoorob, Skocpol, Skocpol and Tervo2020).Footnote 9
Does the activity of such organizations play a role in forging the rural-urban political divide? To investigate this question, we examine the presence of evangelical churches in rural and urban communities by using data from the U.S. Religious Census (ASARB 2020). (We lack sufficient over-time data on other organizations mentioned here.) Religion has loomed especially large in the shifts in partisan coalitions over the past several decades, with more religious people siding with Republicans and those who are more secular siding with Democrats (Margolis Reference Margolis, Lieberman, Mettler and Roberts2021). A key component of this has involved evangelicals as they have increasingly aligned themselves with the GOP, as the party embraced conservative stands on social issues. The Christian Right emerged as a political force in the late 1970s, and its power became evident in the surprise 1980 Reagan landslide, after which Moral Majority leader Rev. Jerry Falwell announced that a “sleeping giant” had been awakened in American politics. Evangelicals played a major role in the transformative 1994 elections by registering and contacting voters and getting them to the polls (Bullock et al. Reference Bullock, MacManus, Mayer and Rozell2019, ch. 5; Wilcox and Robinson Reference Wilcox and Robinson2010, ch. 2). Throughout, it has organized particularly around abortion, gay rights, and schooling, and sought the appointment of conservative judges. Its influence has been evident not only in southern states, but also in the Midwest. When Trump ran in 2016, Christian conservative organizations played a less visible role in the election than in some others, and yet evangelicals themselves strongly supported Trump (Wilcox Reference Wilcox, Rozell and Wilcox2018, 181-183).
Using data from the U.S. Religious Census and the Association of Statisticians of American Religious Bodies (ASARB 2020), we find that the presence of evangelical groups has not shifted much over time; rather they have been disproportionately concentrated in rural areas throughout, by a ratio of nearly 3 to 1.Footnote 10 Their presence in rural counties could be acting as a key organizational venue for conservative politics, and thus contributing to the divide. We expect that once the political economy changed and rural residents felt “left behind,” that subsequently evangelical churches were better able to organize rural voters around issues such as abortion, engaging them in politics. Rural voters affiliated with such churches likely viewed the values of highly educated urbanites to be at odds with their own values, resented their efforts to promote the nationalization of those values, and mobilized against the party that did their bidding. We expect Americans from counties with a higher density of evangelical congregations to be more supportive of the Republican Party, and we expect that this relationship has come to matter more in later periods, after the emergence of political-economic or “left behind” factors.
Racism and Racial Threat
Race and ethnicity have remained defining features of American politics. In 2008, the nation elected its first Black president, but just eight years later, Donald Trump was able to secure the presidency after running a campaign featuring racist and nativist rhetoric. Scholars studying political behavior have found that, in recent decades, white Republicans and Democrats have polarized on racial attitudes, and that, even controlling for other factors, racial resentment – the measure typically used to study anti-Black attitudes—has become an increasingly strong predictor of partisan identification and can influence vote choice (e.g., see Enders and Scott Reference Enders and Scott2019).
How are these attitudes distributed across rurality? Nelsen and Petsko (Reference Nelsen and Petsko2021) found that, as of 2019, anti-Black feelings were strongly associated with rural consciousness—but their cross-sectional analysis begs the question of how these attitudes might have changed over time. Using data from the American National Election Study (ANES 2021) cumulative file, which is geo-coded with respondents’ county of residence, we examine one measure of anti-Black attitudes among rural and urban white people back to 1976. On a scale of 1 to 7, the question asks if “the government in Washington should help improve the economic and social conditions of Black people”; responses range from the “government should make every effort” to “Blacks should help themselves.” While we would ideally have the traditional indicator for racial resentment, the relevant questions are only asked beginning in 1986 (and not for all elections), limiting our ability to track change over our entire time period of interest.Footnote 11 Nevertheless, we think the question here is a good stand-in for racial animus: it picks up on the main elements of racial resentment—namely, the extent to which white people are unsympathetic to Black Americans, and moreover, deem them responsible for racial inequality.Footnote 12 Responses are also strongly correlated with the traditional racial resentment score (r = .66, p = .001).
Figure 6 thus rescales the question on government aid to Black Americans from 0 to 100 for interpretability. High (low) scores indicate support (opposition) for government efforts to help Black Americans. As figure 6 shows, from 1976 to 2008, rural and urban white people tended to score, on average, quite similarly. Beginning in 2012, the two groups began to liberalize on the measure, but urban dwellers more so; by 2020, the difference reached a modest but non-trivial 10 percentage points.Footnote 13
Racial animus by rurality, over time
As racial attitudes have shifted, particularly since the enactment of immigration policy in 1986, the United States has grown more racially and ethnically diverse, including in rural areas (Lichter Reference Lichter2012; Lichter et al. Reference Lichter, Parisi, Taquino and Grice2010). With data from the U.S. Census, figures 7 and 8 graph the changing racial and ethnic demographic profiles of rural and urban counties in recent decades, showing the share of the population identifying as nonwhite and as Hispanic, respectively. As displayed, rural and urban counties featured fairly similar levels of diversity in 1970, but in recent decades, urban counties grew far more diverse. By 2020, as seen in figure 7, over 45% of all urban residents identified as nonwhite. To be clear, rural areas have also diversified significantly over this period, yet as figures 7 and 8 show, the pace of this growth diverged between rural and urban counties, particularly in the 1990s and into the early 2000s. As such, we wonder to what extent rural and urban whites might be responding to these demographic shifts.
Share of population identifying as nonwhite, over time
Share of population identifying as Hispanic, over time
Scholars studying the effect of diversity at the ecological level have offered and found empirical evidence for two competing theories on how these shifts might affect political behavior. The “contact” theory holds that racial attitudes are learned early in life, but can be unlearned through social contact with so-called racial “out-groups” (e.g., Oliver Reference Oliver2010; Oliver and Mendelberg Reference Oliver and Mendelberg2000; Branton and Jones Reference Branton and Jones2005). The “threat” hypothesis anticipates that as the local population becomes more diverse, non-Hispanic whites may feel that their position is threatened by competition for scarce resources, particularly if nonwhites appear to be making political or economic gains. Scholars finding support for the latter theory have shown that it can influence candidate choice (Jardina Reference Jardina2019, 100), and can be activated by economic shocks (Baccini and Weymouth Reference Baccini and Weymouth2021).
How, then, could these developments be contributing to the rural-urban political divide? We expect that issues related to race and ethnicity have sparked backlash politics, particularly in rural areas. We suspect that anti-Black attitudes will play an especially strong role in facilitating support for Republicans among non-Hispanic, rural white people. We furthermore expect that whites, particularly in rural counties, with a higher percentage of Latinos, the ethnic group primarily driving diversity in rural and urban areas, will be more susceptible to threat, and therefore offer greater support for Republican candidates in recent decades. We expect, finally, and most central to our own argument, that these relationships may have intensified following the titanic shifts in the political economy (Baccini and Weymouth Reference Baccini and Weymouth2021), as well as the elections of presidents Obama and Trump—that is, from 2008 onward.
Summing Up
On multiple measures, small differences between rural and urban areas in the 1970s have widened into vast ones, particularly since the 1990s. Rural and urban places, to those who dwell in them, may seem increasingly like two separate worlds.
How might these changes have mattered for how residents of counties vote in presidential elections? We expect that once changes in the political economy carved a deep divide between rural and urban areas, with rural areas relegated to a marginal status, that residents began to shift their voting patterns, with different mechanisms generating change in a sequential process. Although nationalizing trends started by the mid-twentieth century, the political-economic integration of rural and urban areas likely held rural and dwellers together for several decades. But once urban areas diverged from rural ones as the economy changed, those living in counties that had long affiliated with the Democratic Party likely felt increasingly abandoned by it, and the nationalizing dimension became more salient. Looking to the Republicans, they may have recognized a party that was resisting nationalizing trends (including those related to race) that are supported by the growing ranks of well-heeled urbanites, and embracing a posture toward government that tried to deter their influence. In turn, organizations more prevalent in rural areas may have mobilized rural dwellers to emerge as a core constituency of a Republican Party, solidifying a deep new political cleavage.
Methods and Data
To probe empirically the theories outlined in this last section, we employ two central strategies. First, using the county as the unit of analysis, we test the “left behind,” educational gap, and organizational theories by merging county-level presidential vote share data with that measuring various economic, social, and political factors. Here, we use the share of votes going to the Republican presidential candidate as the dependent variable to proxy the extent to which rural and urban counties have diverged politically. In most models, we condition our central independent variables on or interact them with rurality to see if rural and urban areas are responding to such factors similarly or differently.
To explore change over time, we pool the data into three time periods: elections from 1976 to 1988, 1992 to 2004, and 2008 to 2020. Theoretically, these periods have substantive importance. As displayed in figure 1, the first period acts as a baseline, or a period before the rural-urban political divide emerged; the second period covers the time period when the divide began to emerge; and the third period marks its intensification. These periods also track well with social and economic descriptive changes illustrated and discussed above (see figures 3–8). Methodologically, these three pools offer a relatively well-balanced set of models, each encompassing the same number of elections and roughly the same number of observations in each, with one exception.Footnote 14 To detect change over time, we formally test whether regression coefficients differ across models, using the suest command in STATA and Wald tests for equality of coefficients.
We rely on Ordinary Least Squares (OLS) regression, including year and region fixed effects, to generate estimates.Footnote 15 Year fixed effects are included to ensure that any one particular election is not driving our results. Controls for region are included to mitigate concerns about any otherwise unobserved confounders at the regional level. Robust standard errors are clustered at the county-level to address heteroscedasticity.
To test the theories related to racism and racial threat, we make a second empirical move. Because the county-level data used in the models described above necessarily involves the aggregation of individuals into a single estimate, using it as a unit of analysis to study hypotheses related to racism or racial threat poses ecological fallacy problems. As such, we rely on individual-level data retrieved from the American National Election Studies (ANES 2021), running from 1976 to 2020. To examine how place-based factors, such as rurality and level of racial diversity, might impact the political behavior and attitudes of rural and urban dwellers, we secured access to restricted data on respondents’ county of residence from the Inter-university Consortium for Political and Social Research (ICPSR). We then merged each respondent with data discussed later. As with our county-level models, we interact our central independent variables of interest with the rurality of a respondent’s county. To examine change over time, we pool residents into similar time periods as outlined earlier and test for differences between periods of interest. Finally, we restrict all respondents to non-Hispanic whites. We do so because tests related to racial threat lend themselves to examining how the most dominant racial or ethnic group responds to shifting demographics, and as noted in the introduction, there does not appear to be much if any divergence among people of color across the rural-urban divide, at least as far as partisanship and vote choice are concerned.
Dependent variables
In our first set of models examining counties’ political leanings, we use votes cast in each presidential election as the dependent variable. Because we are primarily interested in why rural counties have become more supportive of Republicans, we use the share of all votes cast for the major Republican candidate. We include all votes cast as the denominator to ensure that our measure picks up on those who might be voting for third party, anti-establishment candidates (e.g., Perot), given their potential appeal in rural areas.Footnote 16 Values are scaled from 0 to 1. Data are retrieved from David Leip’s Atlas of U.S. Presidential Elections (Reference Leips2021). In our models testing the racial threat and contact theories, we examine how place and county-level racial and ethnic demographics might shape white rural dwellers’ support for the Republican Party. We operationalize support for Republicans in terms of presidential vote choice, running a series of logistic regressions predicting respondents’ likelihood to vote for Republican candidates.Footnote 17 Voting for a Republican is coded “1” and for a Democrat or any other candidate as “0”.
Independent variables
Rurality. We use the OMB’s definition of rurality, which defines counties in each decade as metro (urban) or non-metro (rural) based on 1) their population density and 2) social and economic integration into core urbanized areas. Relative to other measures of rurality that rely purely on population density (such as the Census Bureau’s), we think this measure thus picks up on a crucial dynamic—the extent to which counties are tied into major cities (USDA 2019). In our data set, each county is thus assigned with a variable so that “1” signifies rural and “0” urban. Granted, using a binary measure can skate over important differences within and across rural and urban areas. Several metro counties, for example, have experienced net population loss, or population shifts within the county over time. Nevertheless, the simplicity of the binary measure makes it helpful in this first attempt to explain the development of the rural-urban divide over time.
Furthermore, because the OMB has changed the classification criteria a number of times, rendering many decades incomparable to others (USDA 2020), following other scholars (e.g., Thiede et al. Reference Thiede, Jaclyn, Brown and Jensen2020) we rely on the 1993 rural-urban distinctions, a year that sits roughly in the middle of our time frame.Footnote 18 However, we note that virtually all of our findings are substantively the same when allowing rurality to vary across decades, by different OMB delineations.Footnote 19
Left Behind. To test the “Left Behind” theory, we use two main variables: population and job growth. Population estimates are taken from the U.S. Census Bureau, while county-level employment numbers are retrieved and calculated from the Bureau of Economic Analysis (BEA), housed within the U.S. Department of Commerce (BEA 2021).Footnote 20 For each measure, we calculate the percentage growth from the previous election year; we then transform and log each value to ensure that outliers are not driving any relationships we find.
Educational gap. To measure educational demographics, we use the share of adults aged 25 and older with at least a four-year degree. Data here are retrieved from the U.S. Census Bureau’s decennial surveys and, when necessary, we used linear interpolation to calculate off-decade, election-year estimates; such interpolation has been shown to perform quite well in producing reliable demographic estimates (Weden et al. Reference Weden, Peterson, Miles and Shih2015).
Organizational mobilization. While we would be keen to examine how numerous organizational developments have potentially influenced the rural-urban divide, due to data constraints we focus on evangelical churches. Data on the presence of evangelical congregations at the county-level are retrieved from the U.S. Religious Census, conducted by the Association of Statisticians of American Religious Bodies (ASARB 2020). The number of congregations for each county is summed and then weighted by the population, yielding a measure that captures the number of congregations per 10,000 residents in a county. As a control for other conservative congregations that may also be acting as a mobilizing force, we include a similar measure for churches in the Catholic denomination from the same U.S. Religious Census.Footnote 21
Racism and racial threat. In the models examining the racial threat and attitudinal theories, we merge ANES respondents with the above-mentioned county-level measures for rurality. To examine anti-Black attitudes, we rely on the question related to government aid explored descriptively earlier. Moreover, because the growing racial diversity in rural and urban areas has largely been driven by the growth of Latinos in American life, we use the presence of people who identify as Hispanic in a respondents’ county as a proxy for racial threat.Footnote 22 In these models, we also include a number of controls at the individual level. These include age, relative income level, educational attainment, and gender. As discussed earlier, analysis here is restricted to non-Hispanic whites.
Demographic trends. Because rural counties tend to be home to older residents who also tend to support the Republican Party at relatively high rates, we control for age by including a measure of the share of the population 65 years and older. To control for racial and ethnic shifts in our first set of county-level models, we calculate the share of each of the population for each county-year that identifies as non-Hispanic white. All data here are secured from the Census Bureau and off-year election estimates are calculated using linear interpolation.
Regression Results: Political Effects of Trends
In conducting our analysis about when and how the trends introduced earlier might be affecting presidential voting in rural versus urban counties, we will be attentive to various patterns. As we have seen, both types of counties have been buffeted by numerous changes over the past several decades and these changes have occurred at different rates and to different extents in rural and urban places. These changes are also associated with political transformation, as we will now see. In some instances, rural and urban people may respond similarly in relation to trends, but if those circumstances are concentrated in rural places, they may be particularly consequential there. We refer to these as “concentrated factors.” Other trends may spur rural and urban people to participate differently at the ballot box, because of the ways those phenomena interact with the circumstances of life in each place. We will refer to these as “distinct factors.” Both dynamics can contribute to the rural-urban political divide, whether or not an interactive relationship exists. We will also be attentive to the timing and sequencing of particular relationships in each area, noting the order in which trends affect politics for each group.
Before testing our theories, we first consider preliminary regression results, a set of simple county-level multivariate models, with Republican vote share as the dependent variable. We begin with these simple associations and differences between the three periods to ensure that rurality does matter above and beyond the demographic characteristics of counties. One might suspect, for example, that the rural-urban divide is a mere artifact of educational or racial polarization; because non-white people and those with higher levels of education are disproportionately concentrated in urban areas, in other words, the rural-urban divide might simply reflect such concentrated differences. To test this, in table A.1 (which, along with each of the other tables, appears in the appendix following the References) we regress Republican vote share on rurality, including county-level controls for levels of education, age, and racial and ethnic diversity. Year and region fixed effects are included.
Rurality is insignificant in our first two periods of interest, but becomes significantly associated with Republican vote share in the third period.Footnote 23 To detect change over time, we formally test to ensure that the coefficient is statistically different from those in the first two models, and it is, implying change over time.Footnote 24 While the coefficient is relatively modest in size, it is worth noting that rural areas hold outsized power through American political institutions, so small differences can be amplified to produce significant changes. Moreover, we underscore that precisely because the rural-urban divide overlaps and intersects with a number of large shifts in political, economic, and social life, rurality’s potential to shape political behavior is likely much greater and articulated in conjunction with other variables, as we show next.
Left Behind
To test the “Left Behind” theory, we turn first to job growth. We start by interacting rurality with logged employment growth, including county-level controls for levels of education, population change, race, and age. Results are listed in table A.2 and depicted in figure 9. We find interactive relationships in all three periods. However, the interactive relationships in the first period is mostly driven by a strong relationship between growth and Republican vote share in urban areas; the rural slope is virtually flat, suggesting very little relationship between job growth and vote share in these counties. In the second period, strikingly, we find that rural areas with low levels of job growth are now more likely to vote Republican, while the opposite remains the case in urban areas. This offers a striking example of “distinct responses” to the same phenomena. In the third period, both rural and urban areas once again exhibit a positive relationship between job growth and Republican vote share.
Place, job growth, and Republican vote share
We now turn to population change. As is shown in table A.3, for the first two periods, even after controlling for job growth, education, race, and age, there is an interactive effect. Yet as displayed in figure 10, that relationship changes over time. In the first period, both rural and urban areas that were growing in population tended to vote Republican. Yet in the second period, as with job growth, while urban areas retained the previous pattern, rural areas reversed: those with declining populations began voting Republican while those with growing populations were more likely to vote for Democrats. We underscore that more than one-fifth of all rural counties between the years 1992 and 2004 lost population. Like job growth, this exemplifies “distinct responses,” with rural and urban people responding differently to the same phenomena. In the third period, we find no interactive relationship, and population growth is once again associated with Republican vote share in both rural and urban counties.
Place, population growth, and Republican vote share
In the 1990s, the political economy of the United States underwent significant transformation, and many urban areas adapted and advanced while rural areas languished. These results indicate that as that occurred, rural voters—many of whom had long viewed the Democratic Party as the one that best represented them—increasingly began to support Republican presidential candidates instead. They did that particularly in areas that hemorrhaged jobs and people.
Educational Gap
Table A.4 shows results for models in which education is interacted with rurality, controlling for race and ethnicity, age, population growth, and job growth. We find an interactive relationship in the first two periods. From 1976 to 1988, education in fact had a stronger relationship with Republican vote choice, particularly in rural areas, as displayed in figure 11. This changed in urban areas in the 1992–2004 period, as counties with more highly educated residents began to support Democratic candidates, while rural counties maintained the earlier in which those featuring lower levels of education favored Democrats. By the third period, however, rural counties reversed course and adopted the same pattern already prevalent in urban counties. We would stress, though, that many urban areas now possess populations with greater levels of formal education than rural areas, while the latter contain concentrated populations of residents lacking college degrees. Through the dynamics of “concentrated factors,” the educational gap divide appears to be contributing to the rural-urban divide.
Place, educational attainment, and Republican vote share
Political economic transformation likely shaped the changing relationship between educational level and political preferences, and nationalizing dynamics are likely at play as well. Notably, changes among urbanites commenced the education realignment. Already in the 1990s and early 2000s, highly educated urban dwellers evidently began to view the Democratic Party as speaking on their behalf. They benefitted from the economic policies the party promoted, seeing their incomes and opportunities rising along with the knowledge economy. The party also increasingly embraced social liberalism, which they favored. Rural dwellers followed suit to realign themselves by educational level, but not until later, in 2008 and beyond. A disproportionate share of them had less education, and they increasingly perceived that party leaders—whom they saw as urban elites (Cramer Reference Cramer2016)—sought to impose their will on the nation. Whether it was on gun control, gay marriage, or environmental policies, they perceived that government—particularly when run by Democrats—was telling them how to live, and they resented it.
Organizational Mobilization
We now turn to examining the effect organizations might be having on the rural-urban divide. We consider the role of evangelical congregations, interacting a population-weighted measure of presence by rurality, with similar controls as introduced earlier, along with one for Catholic congregations. As table A.5 shows, we find statistically significant interactive relationships in all periods, though they differ in size and direction. As displayed in figure 12, the interaction in the first period is largely driven by the mobilizing force evangelical congregations played in urban areas. By the second period, evangelical congregations pushed counties to support Republican candidates in both rural and urban areas, but the relationship remained weaker in rural areas. By the third time period, the rural slope had grown,Footnote 25 and while the slope is smaller than in urban areas, it is important to note that rural areas feature many more congregations, on average, than urban ones.Footnote 26 This is, once again, a “similar response with concentrated effects.” In 2000, for example, the average rural county had roughly 11 congregations per 10,000 people, while the average urban county had just 3.5.
Place, evangelical congregations, and Republican vote share
Here again, urbanites led the way. As social conservatives became ascendant in the Republican Party in the 1980s, they gained support from evangelicals in metropolitan areas, but those in rural areas continued to be loyal to the Democratic Party. That relationship weakened some by the 1990s and early 2000s, as Republican Party officials worked to get gay marriage on the ballot in many states. By 2008 and beyond, people in rural areas with more evangelical congregations became a stronghold in the GOP.
Racism and Racial Threat
We turn now to examining whether the changing racial animus and demographics of rural and urban individuals and communities might have shaped support for the Republican Party. We do so by using individual-survey data from the American National Election Studies Cumulative Time Series file (ANES 2021), and merging place-based data proxying rurality and racial demographics with respondents’ counties of residents. As discussed earlier, because the rural-urban political divide appears to be driven by non-Hispanic whites, we restrict our analysis to them here.
How might racial attitudes shape individual support for the Republican Party among non-Hispanic whites? In logistic regressions we find that racial animus (as proxied by the question about government aid to Black Americans) in all three periods is associated with Republican vote choice, even after controlling for other important covariates, including local racial demographics (see table A.6).Footnote 27 Yet to probe differences across rurality, we interact respondents’ support for government assistance to African Americans with their place of residence in terms of rurality. Results are listed in table A.7 and graphed in figure 13. In period one, we find an interactive relationship, in which both rural and urban dwellers are more (less) likely to vote Republican if they are less (more) supportive of aid to Black Americans, though the relationship is stronger among urban people. In period two, we find no interactive relationship; rural and urban people are both more (less) likely to vote Republican as they exhibit more (less) racial animus by our measure. In the third period, the interaction term is not significant at conventional levels, and rural and urban respondents exhibit similar patterns as in period two. However, while the regression slopes themselves do not differ, note that the predicted probability of voting Republican among rural white people is higher, all else equal, for most values of racial animus. This is particularly so for those values that are most often selected by respondents.Footnote 28 We take this, as well as our descriptive findings, as evidence that racial animus is contributing to the rural-urban divide, particularly in period three.
Place, racial animus, and Republican vote share
What about growing diversity of place? We examine the racial and ethnic threat hypothesis by running similar regressions as detailed previously, but instead interacting rurality with share of the population that identifies as Latinos in an ANES respondents’ county. Results are in table A.8 and predicted probability plots in the online appendix (see figure A.1). In our first period, we find an interactive relationship, in which rural whites were more likely to vote Republican as the share of Latinos increased in their counties of residence, while urban whites were more likely to vote Democrat. In the second period, we find a similar relationship, but one that is weaker in significance and size. In the third period, we find no interactive relationship at all.
What are we to make of these results? On the one hand, the descriptive statistics on anti-Black racial attitudes coupled with the regression results depicted in figure 13 in period three seem to provide evidence that anti-Black attitudes are contributing to the rural-urban divide, at least in more recent elections. At a minimum, the higher concentration of racial animus in rural areas and its association with Republican support (table A.6) itself acts as a “concentrated factor.” On the other hand, we find less support for the racial and ethnic threat hypothesis. Our results suggest that the greater racial and ethnic diversity that emerged in both rural and urban counties in more recent decades is not related to the growth of the rural-urban divide. The tendency of rural whites in areas with more people who identify as Hispanic to vote Republican was highest in the first period, the baseline period, and it was weaker in the next period and then disappeared. While those dynamics of perceived racial threat may also reveal hostility to nationalization, they occurred in an era of less diversity, and were on the downswing by the time the rural-urban divide began to emerge.Footnote 29 In short, to the extent that racism and nativism are contributing to the rural-urban divide, it is at the attitudinal rather than demographic level.
Summing Up: Sequential Polarization
How did the United States’ rural-urban political divide emerge? Certainly numerous changes have roiled rural and urban areas since the late 1970s, turning them into places with increasingly different experiences of American life. Our analysis suggests that there is no one factor that single-handedly drove their politics apart. Instead, a number of forces, in different places at different times, are associated with the emergence of this divide. By disentangling the timing through which each of several changes transformed rural and urban presidential voting, we have uncovered a developmental and sequential process that ensued across several decades.
Following the 1976–1988 baseline period, political-economic tumult roiled rural places in the 1992–2004 period, a harbinger of the rural-urban political divide. Numerous changes transformed the national economy, and as a result, many urban areas ascended economically while rural areas deteriorated. In this period, it was rural dwellers who changed their political behavior, as those living in areas with stagnating populations and economies tended to shift to voting for Republicans. Evidently, they ceased to view the Democratic Party as having their interests at heart, and they moved on. “Distinct factors” among rural Americans offer support for the “left behind” thesis.
In the wake of the dramatic political-economic shifts, the nationalizing dimension became activated and took hold. From 2008 to 2020, the nationalizing dimension became activated as the “educational gap”—which had already in the previous period drawn highly educated urbanites to the Democratic Party—grew prominently associated with the rise of Republican voting among less-educated rural dwellers. Anti-Black racism, disproportionately concentrated in rural areas, also helped contribute to the divide. As Democratic politicians used the power of government to promote broad national rights of citizenship for gay Americans, Black people, immigrants, and those caught up in the criminal justice system, as well as to protect the environment and promote gun control, rural people defined themselves politically in opposition to these efforts. Less so than during the 1960s and 1970s, when the pursuit of rights had encountered ample resistance by urban whites, by the early twenty-first century, resistance was more likely to flow from less populated areas, as they switched their allegiances to Republican candidates.
In addition, organizations became associated with the rise in this new political cleavage, as rural people who lived among a high concentration of evangelical congregations shifted their support to Republicans. Here again, the greater prevalence of this phenomena among rural areas made it particularly consequential, helping Republican leaders to gain rural supporters.
In sum, several factors are associated with the emergence of the rural-urban political divide, each with specific timing and sequencing of effects. In some respects, rural dwellers themselves took the lead on political change and forged away from urban dwellers with similar circumstances. In other ways, urban Americans changed their political behavior first, and then rural Americans followed suit later on. The combination of these trends adds up to what can be termed “sequential polarization,” a two-staged process that drove apart the voting behavior of rural and urban people. The political-economic dimension took center stage in the first phase of this process, as many in areas that experienced being “left behind” abandoned the Democratic Party. It helped give rise to the second phase, in which rural people—no longer tethered to the Democrats—increasingly shifted to the party that resisted the use of national power to pursue a wide array of policy goals. Through each of these dynamics, a deep political cleavage emerged where it had not previously existed, increasingly dividing rural and urban dwellers as inhabitants of two separate political camps.
Conclusion
Throughout most of American history, the two major political parties each found supporters in both rural and urban parts of the nation. In analyzing how the United States’ rural-urban cleavage emerged in recent decades, an analytical framework offered by scholars of comparative politics has proven useful. It has revealed that the divide did not emanate from just one source but from several, and it did not emerge in a single instant but through a series of gradual processes across many years. Several theories offered by scholars of American politics and comparative politics each shed light on one or another aspect of what ensued, but the full development of the rural-urban political divide involves multiple factors arrayed in processes of sequential polarization.
Americans in rural and urban areas of the United States have, particularly since the late 1990s, experienced contemporary life in very different ways. Rural areas, with some exceptions, have experienced local economies that are mostly stagnant or deteriorating and populations in decline; urban areas have undergone considerable change, but it has more typically involved employment growth and population growth, albeit often amid high inequality. Rural areas have seen only modest improvements in the educational attainment of residents, whereas those in urban areas have made greater leaps, producing a larger density of highly educated people, support for the “educational gap” theory. Some rural areas have gained greater racial and ethnic diversity, but at a rate that is dwarfed by many urban areas, in which the population has become far more diverse. In these and other ways, life in rural and urban areas—not so different from each other in the 1970s or 1980s—has diverged sharply.
Now that the United States has developed a rural-urban political divide, it is poised to be particularly consequential for the nation’s politics, more so than in other nations. This is because American political institutions grant extra leverage to less-populated places, most notably through the Electoral College and the U.S. Senate. When those advantages are consolidated in one of the two major political parties, as they are currently, they can permit that party to wield outsized power relative to the votes it receives in elections (Rodden Reference Rodden2019). Such a party may subsequently be able to stack the deck and cement its dominance, even if it represents a minority of citizens.
Several other topics also await further analysis. We have focused mainly on rural dwellers; the reactions of urbanites also deserve further inquiry (though see Rodden Reference Rodden2019). Pinning down the precise mechanisms related to the nationalization dimension—including the relative salience of particular policies (e.g., gun control), the growth and geographical concentration of partisan media, and so on—remains a task to be taken up by future studies. We have not, for example, addressed the efforts of political leaders and organizations to highlight rural-urban disparities as a means to attract supporters. For rural dwellers, worse economic circumstances can provide fodder for grievances and resentment. Particularly if political leaders evoke and organize such sentiments into politics, they may engender social polarization and “us” versus “them” politics (Abramowitz Reference Abramowitz2018; Mason Reference Mason2018; McCoy and Somer Reference McCoy and Somer2019). Scholars should also probe the emergence of the rural-urban political divide in other aspects of U.S. politics, such as in congressional voting and the politics of particular states. A particularly pressing topic is the extent to which this divide contributes to polarization among elected officials. Further, while we have found evidence that anti-Black attitudes matter, other mechanisms that pertain to racism and nativism also require analysis. Possibly as the Democratic Party shifted on trade and other economic issues, the party image—in terms of who it represented—changed in ways that alienated white rural voters. The activation of racial animus and changing party images could have coincided, furthermore, with the rise of conservative media, particularly in rural areas. The focus on rural counties necessarily highlights white rural dwellers, since they are the majority of rural residence, and yet the 25% of rural dwellers who are nonwhite may respond differently in politics (cf. Brown et al., Reference Brown, Jauregui, Mettler and Riveran.d.). Finally, examining additional variation between and within rural and urban areas is a particularly fruitful avenue for future research. Such research could include moving beyond employing a binary measure—as we have here—by placing rurality instead on a continuum (e.g., see Scala and Johnson Reference Scala and Johnson2017). Additional research could also move below the county, which is admittedly a large unit of analysis, to probe for important differences between big cities and surrounding suburbs.
The contemporary rural-urban divide has become a defining feature of American politics, intersecting with a number of forces to generate significant political change. It is fostering polarization and threatening democracy itself. Understanding its sources moving forward will be crucial to bridging the divide and safeguarding democracy.
Supplementary Material
To view supplementary material for this article, please visit http://doi.org/10.1017/S1537592723002918.
Acknowledgments
The authors are thankful to numerous colleagues who generously read earlier drafts and offered insightful comments that helped them improve it. They include David Bateman, Dan Carpenter, Peter Enns, Dan Lichter, Ken Roberts, Alexander Sahn, and Eric Schickler; anonymous reviewers for this journal; and participants in workshops at the University of Wisconsin–Madison and at Cornell University, Princeton University’s 5th Annual Conference on Identity and Inequality, and the First Annual Consortium of the American Political Economy Research Conference. We are also grateful for funding from the New Frontiers grant from Cornell University’s College of Arts and Sciences and helpful staff at the Inter-University Consortium of Political and Social Research at the University of Michigan.
Appendix
Place and presidential voting
Note: Robust standard errors clustered at county level. *p < 0.1; **p < 0.05; ***p < 0.01
Place, employment growth, and presidential voting
Note: Robust standard errors clustered at county level. *p < 0.1; **p < 0.05; ***p < 0.01
Place, population growth, and presidential voting
Note: Robust standard errors clustered at county level. *p < 0.1; **p < 0.05; ***p < 0.01
Place, education, and presidential voting
Note: Robust standard errors clustered at county level. *p < 0.1; **p < 0.05; ***p < 0.01
Place, evangelical congregations, and presidential voting
Note: Robust standard errors clustered at county level. *p < 0.1; **p < 0.05; ***p < 0.01
Racial animus and individual vote choice
*p < 0.05, **p < 0.01, ***p < 0.001
Place, racial animus, and individual vote choice
*p < 0.05, **p < 0.01,***p < 0.001
Place, racial demographics, and individual vote choice
Standard errors in parentheses. *p < 0.05, **p < 0.01, ***p < 0.001
Place, racial demographics, and Republican vote choice
Open access
